Businesses need to be able to anticipate what products their customers will buy so that they can better respond to changing market demands and consumer tastes. The purpose of this study is to employ several machine learning models that can reliably estimate the customer's likelihood of purchasing the product given a customer's profile, transaction date, and other transaction information. This was achieved by training and evaluating different machine learning techniques, such as naive bayes, linear models, deep learning, and decision trees, on a dataset consisting of actual transaction data from three months of sales at a medium-scale grocery store in Bandung. Results indicated that naive bayes performed best as a prediction algorithm, this study shows that data mining can be used to predict grocery store datasets. This research provides insights into how machine learning can be used to improve businesses' ability to anticipate consumer behavior and respond to changing market demands. We also found that demographic factors like age and location, as well as contextual factors like time of week, significantly influenced customers' propensity to buy.